Principal-agent models when the agent's preferences are state-independent (R&R at Games and Economic Behavior)
Abstract: This paper studies a principal-agent model in which the agent's preferences are state-independent. Such preference structures arise in a variety of settings, including Bayesian persuasion, delegation, cheap talk, and probabilistic verification. I first characterize the set of incentive efficient allocations when the principal can commit ex ante. The analysis shows that every incentive efficient payoff pair can be implemented by a binary mechanism that randomizes over at most two alternatives in each state, extending existing results on principal-optimal mechanisms without imposing restrictions on the principal's preferences. I then study implementation when the principal's commitment power is limited. Incentive efficient allocations are generally not implementable through cheap talk. However, I show that they can be implemented through simple default mechanisms that restrict the principal's ex post choice set. The results illustrate how limiting unilateral decision rights can improve information transmission and increase welfare for both parties.
Preventing Strategic Manipulation with Redundant Information (joint with Nicolás Figueroa, Rubén Jara and Aaron Nicholas) (Online Appendix)
Abstract: Information often reaches decision makers through intermediaries with incentives to manipulate what they report. We show that preventing intermediaries from learning the content of transmitted messages through outcome blinding or other forms of conventional encryption need not eliminate these incentives. An intermediary with prior beliefs about the likely content of a blinded message may still benefit from altering it. We propose a simple protocol that augments messages with redundant information before encryption, thereby removing incentives for manipulation while preserving the information available to the decision maker. We establish conditions under which this approach succeeds both when the intermediary's bias is known and when it is unknown. We test the protocol experimentally by assigning participants the roles of electoral agents reporting voting outcomes. We find that even with outcome blinding, participants recognize opportunities to manipulate information and that introducing redundant information successfully reduces the frequency of such manipulation by 41%.
The dynamics of religious conversion: Theory and Evidence (joint with Akwasi Ampofo, Umair Khalil and Laura Panza)
Abstract: We develop a theoretical model of religious conversion in which potential adherents face changing payoff structures, spiritual or instrumental, triggered by the arrival of a new religious movement, the conversion of monarchs, or other shocks that alter the relative returns to conversion. Coupled with non-linear private returns in the share of co-religionists and intergenerational transmission of religious beliefs, our model yields multiple, stable steady-states in the community's religious composition. We use this model to rationalize the spread of Islam in West Africa post-1500. The Atlantic slave trade altered payoffs in favor of Islam because Muslim rulers offered a de facto protection from enslavement, but only to their co-religionists. This created strong incentives for conversion among non-Muslims in regions exposed to slave raiding. Using newly compiled historical data alongside contemporary survey-based measures of Islamic presence, we show that communities under Islamic rule during 1500-1900 and exposed to Atlantic slavery exhibit persistently higher Muslim presence both after the end of slavery and today.
Assignment mechanisms with state-independent preferences and independent types (online appendix)
Abstract: I study a general mechanism design problem without transfers under two assumptions: that agents have state-independent preferences and statistically independent types. I characterize optimal mechanisms for the one-agent and multiple-agent cases, discuss the value of commitment power for the decision maker and prove a comparative statics' result. As an application, I consider the optimal design of peer-review contests when jurors are biased. I argue that quotas, where each academic field is assigned a fixed number of prizes, are not optimal, and that asking reviewers to review applications from other fields may actually increase the expected quality of the prize recipients.
Can impartial peer-review mechanisms replace external reviewers?
Abstract: The increase in the number of academic competitors for grants and conferences makes it unfeasible to continue to rely exclusively on external reviewers for peer evaluation. An alternative is to ask applicants to rate each other using impartial mechanisms which eliminate the incentives to misreport of self-interested applicants. I introduce the reliable subset mechanism and argue that, for most academic contests, this impartial mechanism performs similarly to using external reviewers, provided the applicants are as capable of generating accurate reviews. I then show that external reviewers are rendered unnecessary altogether whenever evaluations of the same applicant are strongly correlated.